Вы находитесь на странице: 1из 53

Quantum Mechanics, Fields and

Symmetries

Giovanni Garberoglio
Interdisciplinary Laboratory for Computational Science (LISC),
ECT*, Fondazione Bruno Kessler.
e-mail: garberoglio@ectstar.eu

Handout for the Academic Year 2017/2018


Revision of November 3, 2017
2
Contents

1 Scattering theory 5
1.1 Classical scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.1.1 Classical cross section . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Exercise 1 Classical cross section of a hard sphere . . . . . . . . . . . . . . . 9
Exercise 2 Classical cross section of the Coulomb potential . . . . . . . . . . 10
1.1.2 Scattering cross section and probability of interaction . . . . . . . . . 10
1.1.3 Kinematics of scattering: laboratory and center-of-mass frames . . . . 11
1.2 Quantum scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.2.1 The Lippmann–Schwinger equation . . . . . . . . . . . . . . . . . . . . 13
1.2.2 Scattering amplitude and cross section . . . . . . . . . . . . . . . . . . 15
1.2.3 Phase shifts and cross section . . . . . . . . . . . . . . . . . . . . . . . 16
1.2.4 The optical theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.2.5 General properties of the phase shifts . . . . . . . . . . . . . . . . . . 20
Exercise 3 Quantum cross section of a hard sphere . . . . . . . . . . . . . . 21
Exercise 4 Quantum cross section of the square well potential . . . . . . . . 26
1.2.6 The Born approximation . . . . . . . . . . . . . . . . . . . . . . . . . . 29
1.2.7 The semiclassical approximation . . . . . . . . . . . . . . . . . . . . . 32
Exercise 5 Semiclassical cross section of a hard sphere . . . . . . . . . . . . 36
1.2.8 Resonances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
1.2.9 Coulomb scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
1.2.10 Scattering of indistinguishable particles . . . . . . . . . . . . . . . . . 41
1.3 General scattering theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
1.3.1 Møller operators and the S matrix . . . . . . . . . . . . . . . . . . . . 45
1.3.2 The S matrix: cross sections, transition probabilities and all that . . . 47
Appendices of Chapter 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
1.A The Wronskian theorem . . . . . . . . . . . . . . . . . . . . . . . . . . 51
1.B Squares of δ functions . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

Bibliography 53

3
4 CONTENTS
Chapter 1

Scattering theory

The main ideas in scattering are shown in Fig. 1.1, where we illustrate the scattering geometry
in the case of a classical particle beam interacting with an external radial potential V (r). If

Figure 1.1: The geometry of a (classical) scattering event. Taken from http://
hyperphysics.phy-astr.gsu.edu/hbase/Nuclear/ruthcross.html.

we send a beam of particles with uniform current density (let’s denote it by I), the deflection
θ will in general depend on the impact parameter (b). Very far from the scattering center
(r much larger than the size of the region where V (r) is sensibly different from zero) one
measures the number of particles per unit time collected as a function θ, that is n(θ). The
differential cross section σ(θ) is defined as the ratio between the number of particles per unit
time dn measured in a solid angle dΩ divided by the incident density I and the solid angle
itself, that is
1 dn 1 dn(θ)
σ(θ) = = . (1.1)
I dΩ I 2π sin θ dθ

5
6 CHAPTER 1. SCATTERING THEORY

Notice that the definition in Eq. (1.1) is very general,∗ and applies to classical, quantum,
electromagnetic, relativistic and even gravitational processes. What changes is the way in
which the function n(θ) is calculated in the different theories. The picture, however, remains
the same. Before diving into the non-relativistic quantum description of the scattering process,
let us analyze it from a classical perspective.

1.1 Classical scattering


Particles of mass m1 and m2 so we have the total mass M = m1 + m2 and the reduced mass
m = m1 m2 /(m1 + m2 ). Let us assume that the potential of interaction depends only on
the modulus of the distances. Hence the center-of-mass motion decouples from the relative
motion. From Newton’s equations

m1 ẍ1 = −∇1 V (|x1 − x2 |) (1.2)


m2 ẍ2 = −∇2 V (|x1 − x2 |) (1.3)

we see that the relative coordinate r = x1 − x2 satisfies the equation

mr̈ = −∇V (r). (1.4)

Angular momentum conservation implies that the motion is limited to a plane (orthogonal to
the angular momentum), where we can use polar coordinates (r, α). The second derivative is
given by†

r = r(t)r̂(t)
dr
= ṙr̂ + rr̂˙
dt
= ṙr̂ + rα̇α̂
d2 r
= r̈r̂ + ṙα̇α̂ + ṙα̇α̂ + rα̈α̂ − rα̇2 r̂
dt2
= r̈ − rα̇2 r̂ + (2ṙα̇ + rα̈) α̂,


from which we get

m r̈ − rα̇2 = −∇V (r)



(1.5)
d 2 
r α̇ = 0. (1.6)
dt
The second equation is the familiar angular momentum conservation, hence we can define
L = mr2 α̇, from which we get

L2
mr̈ = −∇V (r) +
mr3
= −∇Veff (r) (1.7)
∗ dσ
And notice also that almost everybody else calls what we have called σ as dΩ . In the literature σ is the
total cross section, a quantity that we will call Σ below.

One can easily derive all of these equations by remembering that r̂(t) = (cos α(t), sin α(t)).
1.1. CLASSICAL SCATTERING 7

where we have defined the effective potential


L2
Veff (r) = V (r) + . (1.8)
2mr2
The deflection θ = π − ∆α due to scattering with initial energy E and impact parameter
b is now obtained directly by quadratures. Considering that L = mv0 b and that 12 mv02 = E,
together with the energy conservation corresponding to equation (1.7),
1
mv 2 + Veff (r) = E,
2
we get
Z
∆α = α̇ dt
L dt
Z
= dr
mr2 dr
L 1
Z
= dr
mr2 v
√ Z ∞
1
= 2b E p dr (1.9)
2
r0 r E − Veff (r)
where we have inserted a factor of 2 accounting for the (symmetric) incoming and outgoing
part of the trajectory, and we have denoted with r0 the distance of closest approach, defined
by the largest value for which v(r0 ) = 0. Notice that the deflection does not depend on the
mass (as it should), and in fact, using the relation L2 = 2mEb2 , we can write
Z ∞
2b dr
θ(b, E) = π − ∆α = π − r . (1.10)
r0 V (r) b 2
r2 1 − − 2
E r
Examples of the shape of the function θ(b, E) are reported in Fig. 1.2 in the case of the
Lennard-Jones potential   
σ 12  σ 6
V (r) = 4ε − . (1.11)
r r
A few things are worth noticing:
1. The deflection can be negative and have large values. The actual deflection is of course
the number given by Eq. (1.10) modulo π.
2. In general, there is more than one value of the impact parameter b which results in the
same deflection θ.
3. For low energies, there seem to be an asymptote. This is a peculiarity of attractive
potentials. If the energy is small enough there is one particular value of the impact pa-
rameter for which the particles makes an infinite number of orbits
before being scattered.
db
Needless to say, this trajectory is highly unstable. Since dθ → 0 the contribution of

this asymptote to the differential cross section is zero.
4. At higher energies – and again for potentials having an attractive region – the asymptote
in θ(b) becomes a minimum, corresponding to a deflection angle θR . In this case, the
contribution to the differential cross section is infinite, hence cross section presents an
asymptote for the angle θR . This phenomenon is called rainbow scattering.
8 CHAPTER 1. SCATTERING THEORY

180
90

Deflection angle (deg)


0
-90
-180
-270 E=0.1
E=1
-360 E=10
-450
-540
0 1 2 3 4 5
Impact parameter / σ

Figure 1.2: The deflection θ(b, E) as a function of the impact parameter b for the Lennard-
Jones potential (1.11). The energies are measured as a function of ε.

1.1.1 Classical cross section


One can then use Eq. (1.10) to evaluate the differential cross section (1.1). Referring again
to Fig. 1.1 we see that the number of particles per unit time being deflected by θ is just the
same as the number of particles whose impact parameter b produces the given deflection, that
is
dn = I 2πbdb
from which, accounting for the fact that different values of b might produce the same deflection
θ, we have

1 X db
σ(E, θ) = I2π bi
I 2π sin θ dθ i
i
1 X bi
= dθ , (1.12)
sin θ
db i
i

from which we can define a total cross section Σ(E), integrating σ(E, θ) over the whole solid
angle, that is
Z π
Σ(E) = σ(E, θ) 2π sin θ dθ
0
Z πX
b
= 2π dθi dθ, (1.13)
0

i db i

whose meaning is (roughly) the effective area of the scattering center which is blocking the
beam of incoming particles. Notice that sometimes is customary to define the differential
cross section σ(E, θ) so that its integral over θ directly gives the total cross section. ‡ In our

Be careful! As far as I know, there is no accepted convention. You might have noticed that I used mine.
1.1. CLASSICAL SCATTERING 9

notation we have
X db
σ(E, θ) ≡ 2π sin θσ(E, θ) = 2π bi . (1.14)

i
i

If one looks closely at the shape of θ(E, b) in Fig. 1.2, it is not clear that the integral
db
(1.13) converges. In fact for large values of b we have θ → 0 and dθ → 0. For these particles
db
both the numerator (b) and the denominator ( dθ ) appearing in the integrand of (1.13) diverge
for small deflections (that is, large impact parameters). In this case we can evaluate θ(E, b)
using the impulse approximation, that is assuming that the particle makes a straight line and
evaluating the deflection as the ratio
Z ∞
1
θ(E, b) ∼ F⊥ dt (1.15)
mv −∞

where F⊥ is the component of the force orthogonal to the direction of motion. If we assume
V (r) ∼ (a/r)n at large distances, we get
Z ∞
1 na b dt
θ(E, b) ∼ 2 2 (n+1)/2
√ dx
mv −∞ (x + b ) x + b dx
2 2
Z ∞
nab b dξ
=
2E bn+2 −∞ (1 + ξ 2 )n
1
∝ (1.16)
Ebn
from which we get for the cross section

db
σ(E, θ → 0) ∝ b ∝ θ−1/n θ−1/n−1 = θ−2/n−1

which is generally non-integrable for θ → 0 (definitely not so for the Lennard-Jones potential).
Some examples of differential cross sections for the Lennard-Jones potential are reported
in Fig. 1.3.

Exercise 1 Classical cross section of a hard sphere


Calculate the classical differential and total cross section for the hard sphere potential,
that is a potential for which V (x < a) = ∞ and V (x > a) = 0.
Answer of exercise 1
The deflection is non zero only if b < a. In this case the deflection is θ(b) = π−2 arcsin(b/a)
(no dependency on the energy, of course) from which we get
   
π−θ θ
b(θ) = a sin = a cos ,
2 2

and hence
π
a2
Z
Σ = 2π sin(θ) dθ = πa2 ,
0 4
so that the total cross section is, as expected, equal to the geometrical cross sectional area of
a sphere of radius a. Notice that in this case we have σ HS = πa2 sin(θ)/2.
10 CHAPTER 1. SCATTERING THEORY

40

2
E/ε=2

Differential cross section / σ


E/ε=3
30 E/ε=5
0

Deflection angle (deg)


-20
20
-40

-60
10
1 1.25 1.5 1.75 2
Impact parameter / σ

0
0 30 60 90 120 150 180
Deflection angle (deg)

Figure 1.3: The classical differential cross section (1.14) of the Lennard-Jones potential for
three values of the energy. The inset shows the deflection angle as a function of the impact
parameter. Notice the divergences at the rainbow angles.

Exercise 2 Classical cross section of the Coulomb potential


Calculate the differential cross section of the Coulomb potential in classical mechanics.
Answer of exercise 2
The most elegant way is to use the Lenz vector, which, for the Coulomb potential in the
form V = Ze2 /r is
A = p ∧ L + mZe2 r̂,
and points towards the periapsis. For the initial condition far from the scattering center the
vectors p ∧ L and r̂ are orthogonal. Since |p ∧ L| = mv · mvb = 2Eb, one has
 
2Eb
θ = π − 2 arctan ,
Ze2
that is
Ze2
 
θ
b= cot ,
2E 2
from which we can get the differential cross section (1.12)

Z 2 e4
σ(E, θ) = . (1.17)
16E 2 sin4 (θ/2)
Notice that the total cross section turn out to be infinite, again because of a divergence at
small deflections.

1.1.2 Scattering cross section and probability of interaction


The total number of scattered particles in a time ∆t from one scattering center is

Nsc = IΣ∆t,
1.1. CLASSICAL SCATTERING 11

but the total number of incoming scattered particles in the same time is

Σ
Nsc = Ninc ,
S

so that we recover the usual interpretation of the scattering cross section as proportional to
the probability of being scattered. The proportionality factor is the inverse cross section of
the beam, of course.

1.1.3 Kinematics of scattering: laboratory and center-of-mass frames


For the sake of simplicity, we have been considering the simplest possible situation, that is
the scattering in relative coordinates. Equivalently, we have been considering the reference
system where the center of mass of the scattering pair is at rest. In practice, one works
in a laboratory setting where particles are sent towards a fixed target (e.g., the original
Rutherford experiments). In this section we will briefly describe how quantities measured in
the laboratory frame are related to the ones used in the description so far.
The intensity I is actually the density of “incoming particles” (ρ1 ) multiplied by the
relative velocity between the scattered particles and the scattering center. In general if n is
the number of scattering events observed per unit time, I is the product of the densities of
both reactants and their relative velocity. As such, it is independent on the frame of reference.
The number of scattered particles per unit time, dn in Eq. (1.1), is also clearly independent
on the choice of the reference frame.
The scattering angle θ, however, does depend on the frame of reference. Let us consider
particles of mass m1 impinging with velocity v1 upon particles of mass m2 at rest in the
laboratory frame. The velocity of the center of mass with respect of the laboratory frame is
then
m1 v1
v=
m1 + m2
hence the velocity of particle 1 in the center of mass system is
m 2 v1
v10 = v1 − v = .
m1 + m2

In this reference system, particle 1 only changes the direction of motion after scattering,
making an angle θ with the initial direction. Back in the lab frame the velocities of particle
1 after the scattering will be

vx = v10 cos θ + v
vy = v10 sin θ

so that the scattering angle in the lab frame, θL , will be

v10 sin θ
tan θL =
v1 cos θ + mm1 +m
0 1 v1
2
sin θ
= m1
cos θ + m 2
12 CHAPTER 1. SCATTERING THEORY

which implies, from cos2 θ = 1/(1 + tan2 θ), and defining τ = m1 /m2 ,
cos θ + τ
cos θL =
(1 + 2τ cos θ + τ 2 )1/2
d cos θL 1 + τ cos θ
=
d cos θ (1 + 2τ cos θ + τ 2 )3/2
and, finally !
dn dn (1 + 2τ cos θ + τ 2 )3/2
σL = = =σ ,
IdΩL I 2πd cos θL |1 + τ cos θ|
which gives a relation between the differential cross section in the center-of-mass frame and
that in the laboratory frame.

1.2 Quantum scattering


How good is classical mechanics in explaining actual experimental data on particle scattering?
There are two conflicting evidences. First of all, the early Rutherford experiments on α-
particles scattered by matter showed that the differential cross section could be very well
fitted with the Coulomb expression (1.17).
On the other hand, atomic beam experiments paint a different picture. Not only are the
total cross sections finite, but the differential cross sections show a distinct diffraction pattern.
This is evident in Fig. 1.4 where we report experimental measurements of the differential cross
sections for H–Ar scattering [BDT+ 76] and actual data on the total cross sections measured
for the scattering of hydrogen atoms on noble gases [TWW79].
Not only there is no sign of the expected divergence at zero scattering angles that could
be expected on the basis of classical considerations (although measurements in this regime
are quite hard because of the presence of the incident beam that pass through the scattering
region without interacting), but the total cross sections are finite and show a very pronounced
dependence on the incoming-beam energy, with a peculiar sequence of maxima and minima.
Also, experiments show a significant isotopic dependence on the differential cross section
(see the left-hand panel of Fig. 1.4). This is in contrast with the classical result for which
the deflection, and hence the cross sections, should not depend on the mass but only on the
energy – cfr. Eq. (1.10) – and that was fixed at 67 meV for both hydrogen and deuterium
in Ref. [BDT+ 76]. In general, the presence of isotopic effects is something that classical
mechanics has difficulties explaining.
In fact, scattering processes at the atomic and subatomic level are yet another indication
that something is not classical down below. In the following, we will see how quantum
mechanics will enable us to obtain a quantitative understanding of all the features observed in
the experimental data, that is diffraction-like differential cross sections, finite cross sections,
and their non-monotonic dependence on the beam energy. In a classical picture, all the trouble
comes for the small-angle deflections, for which the differential cross section diverges.
In fact, a heuristic application of the uncertainty principle, shows that this is indeed the
region where one could expect quantum mechanical effects to be the most important. [Chi96,
Chapter 2]. If we consider an incoming ring of particles of width db (see Fig. 1.1), then by
Heisenberg principle we must associate an uncertainty in the orthogonal momentum
~
dp⊥ ∼ ,
db
1.2. QUANTUM SCATTERING 13

Figure 1.4: Left: Differential cross section for H–Ar scattering, taken from Ref. [BDT+ 76].
Right: Total cross sections for the scattering of H with noble gases, taken from Ref. [TWW79].

which in turn results in an uncertainty of the incoming angle

dp⊥ ~ ~
dθ = =  .
mv mv db mvb
Quantum mechanical effect will be appreciable when this uncertainty is greater than the
value of the scattering angle which, from Eq. (1.16), has the form θ ∼ Kb−n in the case
of potentials with long-distance tails. Therefore a sufficient condition for the presence of
significant quantum effects might be written as
 1/(n−1)
~ mvK
 Kb−n ⇒ b  ,
mbv ~

which indeed corresponds to small-angle scattering. Let us then set up to the task of solving
Schrödinger equation for a scattering process.

1.2.1 The Lippmann–Schwinger equation


[Wei15,
In actual experiments a continuous beam of particles is directed onto a target and the scattered §7.2]
particles are measured. We assume that the region of interaction is microscopic (nanometers,
14 CHAPTER 1. SCATTERING THEORY

in the case of atoms), and that measurements are perfomed on a macroscopic scale (meters,
in the case of actual laboratories). Analogously to classical mechanics, even in quantum
mechanics one can separate the center-of-mass and relative coordinates, so we will assume
without loss of generality that scattering occurs on a fixed potential V (r) which is spherically
symmetric, and we put the symmetric center at the origin of our coordinate system. The
separation of scales between microscopic and macroscopic world implies that lim V (r) = 0.
r→∞
We will see in the following more restrictive conditions on the form of V (r). Additionally, we
also assume that a stationary state has been reached.
Hence, we would like to find stationary solutions of the Schrödinger equation which de-
scribe a freely moving beam of particles at large distances from the scattering region, super-
imposed with a scattered beam coming from the interaction zone. Differently to bound state
problems, in this case the energy of the stationary states must be positive.
This is most conventiently done by rewriting the differential Schrödinger equation into
an integral form, using the theory of Green functions. The resulting equation is named after
Lippmann and Schwinger. Formally one can write for the scattering state |ψi

(E − H0 )|ψi = V |ψi, (1.18)

which becomes
|ψi = |ϕi + (E − H0 + i)−1 V |ψi, (1.19)

where |ϕi is a solution of H0 |ϕi = E|ϕi. Eq. (1.19) is the venerable Lippmann–Schwinger
equation. Notice that it represents the time-independent scattering state |ψi as the superpo-
sition of an incoming free-particle state |ϕi and a purely scattered state (E − H0 + i)−1 V |ψi.
Notice that the sign of +i has been introduced to make the scattered state an outgoing
wave. In fact, the free Schrödinger equation is a second order equation, and can represent
waves that are both going towards and from the scattering center. Only the latter possibility
is the one realized in an actual experiment.
To see it better, let us project Eq. (1.19) on a set of position eigenstates and let us define
ψ(r) = hr|ψi. We also assume that the incoming free-particle state is a plane wave with
momentum k and energy E = ~2 k 2 /(2m), that is ϕ(r) = hr|ϕi = eik·r .§ The expression of
the scattered state becomes
Z
−1
hr|(E − H0 + i) V |ψi = d3 qd3 r0 hr|qihq|(E − H0 + i)−1 |r0 ihr0 |V |ψi
0
2m eiq·(r−r ) 1
Z
3 0
= 3
d qd r 2 V (r0 )ψ(r0 )
~ (2π)3 k 2 − q 2 + i
Z ∞
2m sin(q|r − r0 |) V (r0 )ψ(r0 )
Z
3 0
= d r dq 4πq 2
0 (2π)3 ~2 q|r − r0 | k 2 − q 2 + i
2m ik|r−r0 |
3 0 e
Z
= − 2 d r V (r0 )ψ(r0 ),
~ 4π|r − r0 |

§
If you are wondering about the normalization, consider that Schrödinger equation is linear and that we
are going to evaluate observables which are independent on the normalization of the incoming state.
1.2. QUANTUM SCATTERING 15

where we have used


0
0 eiq·(x−x )
hx|qihq|x i =
(2π)3
Z ∞
sin(q|r − r0 |)
Z
3 iq·|r−r0 |
d qe = 4π dq q
0 |r − r0 |
Z ∞ 0
4π eiq|r−r |
= dq q
2i −∞ |r − r0 |
1 1
2 2
= − ,
k − q + i (q − k − i)(q + k + i)

and the theorems of residues, where we closed the integration counter in the upper complex
0
plane because of the presence of the eiq|r−r | term. As anticipated, the presence of the +i
term selects the q = k + i pole resulting in the appropriate (that is, outgoing) Green’s
function.¶

1.2.2 Scattering amplitude and cross section


Let us now consider the fact that we are interested in the behavior of the scattered state far
from the scattering center. The Lippmann–Schwinger equation (1.19) in coordinate space is
0
2m eik|r−r |
Z
ψk (r) = e ik·r
− 2 d3 r0 V (r0 )ψk (r0 ), (1.20)
~ 4π|r − r0 |

where we have added a suffix k to ψ(r) to indicate that the solution does depend parametri-
cally on the direction of the incoming wave.
Using spherical coordinates centered at the origin of the potential and defining r = |r|,
r̂ = r/r and using the expansion valid for r  r0 = |r0 |

r · r0
|r − r0 | ' r − = r − r̂ · r0 ,
r
we can write
m eikr
Z
0 0 0
ψk (r ) = e ik·r
− 2
d3 r0 e−ik ·r V (r0 )ψk (r0 ) (1.21)
2π~ r
eikr
≡ eik·r + f (k, k0 ) , (1.22)
r
where we have defined k0 = kr̂ (the direction of observation) and the scattering amplitude

m
Z
0 0 0
f (k, k ) = − 2
d3 r0 e−ik ·r V (r0 )ψk (r0 ), (1.23)
2π~

which depends on the solution ψ(r0 ) and is function of the modulus k of the incident wavevec-
tor, as well as of the angle of observation cos θ = k · r/(kr) = k · k0 /k 2 . The form of the
scattering wavefunction (1.22) enables one to directly derive the expression of the cross section

The kinetic energy term in the Schrödinger equation is a second order term, hence there are two Green’s
functions which, in radial coordinates, represent outgoing and ingoing spherical waves.
16 CHAPTER 1. SCATTERING THEORY

as a function of f (k, k0 ) = f (k, θ), where we have made explicit the fact that the scattering
is elastic (|k| = |k0 |), and that the system possesses cylindrical symmetry.
The differential cross section can be calculated using the definition (1.1) and noting that
in the quantum mechanical case we have
~k
I = |jin | = (1.24)
m
dn
= r̂ · jout r2 (1.25)
dΩ
~k 2
= f (k, k0 ) , (1.26)
m

where jout is the current corresponding to the scattered wavefunction k

eikr
ψout = f (k, k0 ) ,
r
from which we obtain the expression

σ(E, Ω) = |f (k, θ)|2 , (1.27)

which relates the scattering cross section to the angular behavior of the scattered (part of
the) wavefunction. Unfortunately, the scattering amplitude f (k, θ) is still implicitly defined
by Eqs. (1.22) and (1.23).

1.2.3 Phase shifts and cross section


[Wei15,
§7.5] Let us now use Eq. (1.22) to express f (k, θ) and hence the cross section as a function of other
quantities, the phase shifts, that are more amenable to numerical calculation. Since we are
working in spherical coordinates centered in the origin of the potential and we have complete
symmetry along the φ angle, this latter does not compare in the equations. Hence, we can
write (1.22) as
eikr
ψk (r, cos θ) = eikr cos θ + f (k, cos θ) , (1.28)
r
and recall that k is fixed by the external conditions (that is, the energy E of the incoming
beam). The idea is to expand both sides of Eq. (1.28) into a suitable set of basis functions in
spherical coordinates (that is, r and cos θ).

Free-particle wavefunction in radial coordinates. In radial coordinates the


expression of the laplacian is

1 ∂2 L2
∇2 · = 2
(r·) + ·,
r ∂r 2mr2
and the free-particle eigenfunction factorizes as

ψk (r, θ, φ) = χ(r)Ylm (θ, φ),


k
Since we are interested only in the radial part of the current, we might just
 consider the
 radial part of
~ ∗ ∂ψ(r)
the momentum operator in spherical coordinates, hence the current is j = Im ψ (r) r̂.
m ∂r
1.2. QUANTUM SCATTERING 17

where the radial part is determined by


~2 1 d2 ~l(l + 1)
− 2
(rχ) + χ = Eχ.
2m r dr 2mr2
There are two linearly independent solutions of the previous equation, usually
called jl (kr) and nl (kr). The first one has the properties
(kr)l

 ∼ (2l + 1)!! kr  1



jl (kr) ∼1 kr ∼ l (1.29)
sin(kr − πl/2)


 ∼

kr  l
kr
where the last equation means that one has to consider a region where centrifugal
potential is much smaller than the kinetic energy, that is
~l(l + 1) ~k 2
 ,
2mr2 2m
p
which is equivalent to kr  l(l + 1). The other solution has the properties
  l+1
 ∼ − (2l + 1)!! 1

kr  1

nl (kr) = 2l + 1 kr (1.30)
 ∼ − cos(kr − πl/2)

 kr  l
kr

Recalling the meaning of the Legendre polynomials Pl (cos θ) we can immediately write
X
f (k, cos θ) = fl (k)Pl (cos θ). (1.31)
l
The expansion of the free-particle wave is more complicated, but it has the structure
X
eikr cos θ = il (2l + 1)jl (kr)Pl (cos θ)
l
X sin(kr − πl/2)
= il (2l + 1) Pl (cos θ). (1.32)
r→∞ kr
l

The nl (kr) functions do not contribute to the expansion because they would diverge at the
origin, where the exponential is regular. Notice the presence of the Legendre polynomials:
since the incoming wavefunction does not depend on the angle φ because of our choice of
coordinates, only the m = 0 spherical harmonics contribute to the expansion in (1.32), and
Yl0 (θ) is proportional to the Legendre polynomial Pl (cos θ).
However, both spherical Bessel functions appear in the asymptotic expansion of the left-
hand side of Eq. (1.28). In this case the r → 0 behavior does not preclude the appearance of
nl (kr) because it the r → 0 limit the presence of a potential results in its amplitude going to
zero. Hence we can write
X
ψk (r, cos θ) = Al Pl (cos θ) [cos δl (k) jl (kr) − sin δl (k) nl (kr)] (1.33)
rλ
l
X Al Pl (cos θ)
= [cos δl (k) sin(kr − πl/2) + sin δl (k) cos(kr − πl/2)]
kr1 kr
l
X Al Pl (cos θ)
= sin(kr − πl/2 + δl (k)), (1.34)
kr
l
18 CHAPTER 1. SCATTERING THEORY

where λ is a typical length scale of the interaction potential such that v(r  λ) ∼ 0. We
immediately recognize the meaning of the phase shifts δl (k) as the phase difference in the
asymptotic regime between the scattered wavefunction ψk (r, cos θ) and the free-particle in-
coming wavefunction. Notice that the values of the phase shifts depends both on the angular
momentum and the energy (or, as it is the same, the wavevector k). In the following, we will
sometime drop the explicit dependence on k to avoid cumbersome notation.
Due to the presence of the Legendre polynomials, Eqs. (1.31), (1.32) and (1.34) must be
fulfilled for each value of l. This gives two unknowns (Al and δl ) and two conditions (the
coefficients of jl (kr) and nl (kr) – or the coefficients of eikr and e−ikr for kr  1 and r  λ –
must be equal to to their linear independence). A bit of algebra results in

2l + 1
fl = sin δl eiδl (1.35)
k
Al = il (2l + 1)eiδl (1.36)

from which we get



X 2l + 1
f (k, cos θ) = eiδl (k) sin δl (k) Pl (cos θ) (1.37)
k
l=0
X∞
≡ fl (k) (1.38)
l=0


4π X
Z
Σ(E) = σ(Ω, E)dΩ = 2 (2l + 1) sin2 δl (k) (1.39)
k
l=0

where we used the fact that


1
2
Z
Pl (x)Pl0 (x) dx = δl,l0 .
−1 (2l + 1)

Notice that if we could somehow (= numerically!) evaluate the radial part of the scattering
wavefunction (1.33) for a given angular momentum, that is the functions ϕl (r) given by

X
ψk (r, cos θ) = ϕl (r)Pl (cos θ),
l=0

in the r > λ region, then we could get the phase shifts directly from the values of ϕl (r) in
two points r1 and r2 from

ϕl (r2 )
K = (1.40)
ϕl (r1 )
jl (kr2 ) − Kjl (kr1 )
tan δl (k) = . (1.41)
nl (kr2 ) − Knl (kr1 )
P∞
Notice that the very important Eq. (1.39), can be written as Σ(E) = l=0 σl , where

4π(2l + 1)
σl (E) = sin2 δl (E),
k2
1.2. QUANTUM SCATTERING 19

is the contribution of the l-th partial wave to the total cross section, which is limited by
4π(2l+1)/k 2 . For some specific energies it might happen that δl (E) = π/2 and the l-th partial
wave gives the largest contribution to the cross section. As we will see below, δl (E) ∼ π/2
usually happens in a narrow range of energies, resulting in a pronounced peak in the cross
section (on top of a background given by contribution of all the other angular momentum
components), called resonance, which we will discuss better in a subsequent section.

1.2.4 The optical theorem


[Wei15,
A very important relation comes from considering the value of the scattering amplitude (1.37) §7.3]
in the forward direction (θ = 0). Since the Legendre polynomials Pl (cos θ) satisfy the relation
Pl (1) = 1, one has

1X k
Im [f (k, θ = 0)] = (2l + 1) sin2 δl = Σ(E),
k 4π
l

where we have used the result of Eq. (1.39). This is the well known optical theorem relating
the total cross section to the imaginary part of the forward scattering amplitude

Σ(E) = Im [f (k, θ = 0)] . (1.42)
k
The physical meaning of this relation is related to the fact that the interference of the
incoming wave and the diffracted wave in the forward direction results in a decreased am-
plitude of the total scattering wavefunction, which corresponds to the fact that some of the
particles are being indeed scattered. To see this, let us consider the form of the scattering
wavefunction (1.33) and let us calculate the flux of the current j on a large sphere around the
scattering center. From the continuity equation of the Schrödinger equation one must have
∇ · j = 0. The radial part current is given by
 
~ ∗ ∂
r̂ · j = Im ψ ψ
m ∂r
−ikr
  
~ −ikr cos θ ∗e ikr cos θ f ik ikr f ikr
= Im e +f ik cos θe + e − 2e . (1.43)
m r r r

After performing the products, we recognize one term ik cos θ which correspond to the current
of the incoming wave eik·r (whose flux integrates to zero, of course) as well as the term

~ 2k
r̂ · jscat = |f | 2
m r
which is the current associated with the scattered wavefunction, resulting in a total positive
flux when integrated on any sphere around the scattering center
Z Z
2 ~k ~k
r r̂ · jscat dΩ = |f (k, θ)|2 dΩ = Σ(E), (1.44)
m m

There is also a purely real term O(r−3 ) which does not count, and a term O(r−2 ) which
would not contribute to the flux (see below). The part that remains to be evaluated is due
to the interference between the incoming wavefunction and the scattered wavefunction. Since
20 CHAPTER 1. SCATTERING THEORY

the total probability is conserved, this must produce a negative flux on spheres around the
scattering center which compensates the flux (1.44) associated to the scattered. It reads

f ik ikr(1−cos θ) f ∗ ik cos θ −ikr(1−cos θ)


 
~
r̂ · jint = Im e + e ,
m r r
from which we see that unless cos θ ∼ 1 is a highly oscillating function of r. Indeed, it
averages to zero for θ 6= 0 once we integrate the contributions over a series of wavevectors
with spread ∆k [Joa84]. So, this part of the current is concentrated on a small cone in the
forward direction. Assuming that the scattering amplitude is slowly varying we can set their
value to the forward direction. The integration of the phase factor gives then
Z 1
1  −ikr  i
d cos θ eikr(1−cos θ) = eikr e − e−ikr cos δ ∼ + oscillating terms,
cos δ −ikr kr
from which we can see that
Z
~ ~
r2 r̂ · jint 2πd cos θ = 2π Im [−f (θ = 0) + f ∗ (θ = 0)] = (−4π)Im [f (θ = 0)] (1.45)
m m
which gives a contribution to the forward flux that must compensate the flux of the scattered
particles. Since the sum of (1.44) and (1.45) must be zero, one obtains the optical theorem
(1.42). Notice also that the integration on cos θ of a term of the form eikr(1−cos θ) introduces a
factor 1/r. When this result is applied to the O(1/r2 ) term in Eq. (1.43), it gets transformed
in a O(1/r3 ) term, whose flux disappear at large distances.
One direct consequence of the optical theorem provides interesting information on the
structure of f (k, θ) at high energies. Since the total cross section Σ(k) is obtained as an
angular integral of f (k, θ), there must be a solid angle ∆Ω so that

1 1 k 2 Σ2 (k)
Σ(k) > |f (k, θ = 0)|2 ∆Ω > |Im[f (k, θ = 0)]|2 ∆Ω = ∆Ω,
2 2 32π 2
hence, the aperture ∆Ω has to be a decreasing function of the energy, since the previous
equation gives
32π 2
∆Ω < 2 , (1.46)
k Σ(k)
where we have assumed that (as observed) the cross section does converge to some definite
value at high energies. This pronounced peak of the scattering probability in the forward
direction is called the diffraction peak.

1.2.5 General properties of the phase shifts


[Wei15,
§7.5] A lot of interesting properties of the phase shifts comes from the formula (1.41). For theo-
retical purposes, it is probably most useful to express them as a function of a single variable
r0 = r1 , by expressing r2 = r0 + δr and taking the limit δr → 0. In this way we obtain
ϕ(r0 + δr) ϕ0 (r0 )δr
K= ∼1+ ≡ 1 + D δr,
ϕ(r0 ) ϕ(r0 )
and, from
jl (k(r0 + δr)) ∼ jl (kr0 ) + kjl0 (kr0 )δr,
1.2. QUANTUM SCATTERING 21

we get
kjl0 (kr0 ) − D jl (kr0 )
tan δl (k) = . (1.47)
kn0l (kr0 ) − D nl (kr0 )
where we still have that the argument has to be evaluated at a point r0 where V (r0 ) ∼ 0.
However, in the low-energy limit where k → 0 the phase shifts are determined by the behavior
of the spherical Bessel functions near the origin – see Eqs.(1.29) and (1.30) – which results in

tan δl (k → 0) ∼ A (kr0 )2l+1 ,

from which we see that only s-wave scattering (l = 0) is expected to contribute. In the low-
energy limit, there might be specific energies where we have A = 0 (this depends of course
on the form of V (r)). If this happens, one observes a very small scattering cross section; this
effect is named after Ramsauer and Townsend, who discovered it for electrons diffusing in
gases way before a quantum theory of scattering was developed.
The previous equation shows that in the low-energy limit one can express the tangent of
the s-wave phase shift as
tan δ0 (k) ∼ −kas , (1.48)
k→0

where the quantity as is called the scattering length. The minus sign in the definition (1.48)
is conventional and such that as turns out to be positive for repulsive potentials and negative
for attractive ones, as we will see below. Notice that in the low-energy limit the differential
cross section becomes a constant, σ = a2s and the total cross section is Σ = 4πa2s .
Conversely, we can consider what Eq. (1.47) implies in the high-energy limit. The general
shape of the spherical Bessel functions is such that both are very close to zero when kr < l.
So, if we assume that the potential has a range r0 , where we can safely evaluate (1.47), then all
the components of the scattering wavefunction with angular momentum greater than kr0 will
not contribute to scattering. In a semiclassical perspective, the centrifugal barrier is so high
that it prevents the particle to enter the region where V (r) 6= 0. Hence, at high energies, we
expect contribution to the cross section (1.39) from all the phase shifts with l < lmax ∼ kr0 .

Exercise 3 Quantum cross section of a hard sphere


Calculate the differential and total cross section for quantum scattering on a hard sphere.
Answer of exercise 3
The spherical components of the scattering wavefunction, ul (r), are zero for r < a for
any value of l. Outside the hard sphere, we have a free-particle solution for any angular
momentum, but both the spherical Bessel functions are present. The coefficients of the
superposition – and hence the phase shifts – determined by ul (a) = 0, hence from (1.33)
tan δl = jl (ka)/nl (ka) which gives

jl (ka)2
sin2 δl = .
jl (ka)2 + nl (ka)2

In the low-energy limit only s-wave scattering counts and the have tan δ0 → ka, hence

Σ(E) = (ka)2 → 4πa2
ka→0 k2
22 CHAPTER 1. SCATTERING THEORY

which is four times the classical value. In the opposite high-energy limit (E → ∞, that is
ka → ∞) one can use the large r limit of the spherical Bessel functions – Eqs. (1.29) and
(1.30) – getting sin2 δl ∼ sin2 (ka − πl/2) hence
ka
4π X
Σ(E) = (2l + 1) sin2 (ka − πl/2),
ka→∞ k2
l=0

with the upper limit of sum determined by the same arguments given in the previous section.
To evaluate the expression, we can see the for even l the addend is sin2 (ka), while for odd
l is cos2 (ka). We can sum over the odd and even parts defining l = 2n and l = (2n + 1)
(n = 0, 1, 2, . . .) obtaining
(ka)/2
 4π 1 ka 2
 
4π X  2 2
Σ(E → ∞) ∼ 2 4n + sin (ka) + 2 cos (ka) ∼ 2 4 → 2πa2 ,
k k 2 2
n=0

which is twice the classical value. The reason why the quantum cross section does not converge

Quantum hard-sphere differential xs


ka = 0.1 ka = 1
1.0 2.0 Quantum
0.8 Classical
1.5
0.6 1.0
0.4 0.5
0 1 2 3 0 1 2 3
ka = 10 ka = 100
103
101 102
101
100
100
10 2 10 1 100 10 2 10 1 100

Figure 1.5: The differential cross section for scattering of a quantum particle from the hard-
sphere potential (continuous line) in units of a2 , compared to the results of classical scattering.

to the classical one at high energies can be seen from Fig. 1.5. At small values of ka the
differential cross section is almost constant and four times as large as the classical counterpart.
No problem here; this is diffraction of a wave, so we can expect some differences with respect
to an inpinging beam of classical point particles. When the energy of the beam is increased,
one sees that the value of the differential cross section at high angles rapidly converges to the
classical one. However, the differential cross section at small angles is characterized by a sharp
diffraction peak, whose width decreases at higher energy by virtue of the optical theorem,
cfr. Eq. (1.46). The intensity, however, increases, so that the integral of the diffraction peak
remains finite. The presence of this peak is responsible for the cross section being twice than
the classical one. In fact, scattering at all energies is a purely ondulatory phenomenon; it is
1.2. QUANTUM SCATTERING 23

Figure 1.6: An actual picture of the Arago spot. Taken from the webpage of Thomas J. Bauer
at the University of Wellesley.

due to the fact that the boundary of the potential is sharp and hence the incoming waves
always have sum coherently in the forward direction, therefore increasing the cross section.
This property is the quantum mechanical counterpart of the Arago spot (also named after
Poisson or Fresnel, see Fig. 1.6), whose observation contributed to establish the wavelike
nature of light.
So, how comes that in our classical world we measure the total cross section of a sphere as
being πa2 ? Well, there is one assumption that breaks down, and that assumption is that the
incident wave is a plane wave, that is a wave of cross sectional area is much larger than the
geometrical cross sectional area of the hard sphere. Because of this reason, the incident wave
can “interfere with itself” and produce the peak in the forward direction. Once we scatter
(quantum) objects whose lateral spread of the wavefunction is less than a, the diffraction
peak would disappear and the total cross section of would become πa2 .

Some interesting relations involving the phase shifts can be obtained using the Wronskian
theorem (see appendix 1.A) [Mes99]. In our case, the scattering wavefunctions are solutions
of  2  
d 2m l(l + 1)
+ E− V − ul (r) = 0, (1.49)
dr2 ~ r2
with the boundary conditions

ul (r → 0) = 0 (1.50)
 
πl
ul (r → ∞) ∝ sin kx − + δl . (1.51)
2

If we now consider two solutions of (1.49) ul and ûl with two potentials V and V̂ , respec-
24 CHAPTER 1. SCATTERING THEORY

tively, we have the the Wronskian has the limits

W (ul , ûl ) → 0 (1.52)


r→0
W (ul , ûl ) → k sin(δl − δ̂l ) (1.53)
r→∞

hence, from the Wronskian theorem,



2m
Z  
sin(δl − δ̂l ) = − 2 ûl V − V̂ ul dr, (1.54)
~ k 0

from which we see that if we consider a small perturbation of a potential ∆V = V − V̂ , then


the variation of the phase shift is

2m ∞ 2
Z
∆δl = − 2 ul ∆V dr,
~ k 0

where we have neglected the effect of the variation of the potential on the wavefunction
(which would contribute to a higher-order term). From this we can see that “repulsive”
(= generally positive) potentials have negative phase shifts, while “attractive” (= generally
negative) potentials have positive phase shifts. This result justifies the definition (1.48) of
the scattering length.

Effective range theory. We can use the Wronskian theorem at different energies
(but the same potential) to generalize the relation (1.48) for the s-wave phase shift.
In fact, given two s-wave solutions of (1.49) of energy E and Ê (this latter close
to zero) – denoted by u = u(k, r) and û = u(k̂, r), respectively – the Wronskian
theorem provides
b
2m
Z
b
W (u, û)|a = 2 (E − Ê) u(k, r)û(k̂, r) dr.
~ a

We have the following properties of u and û:

u(k, r = 0) = 0
u(k, r → ∞) ∼ cos(kr) + cot δ0 (k) sin(kr)
û(k̂, r = 0) = 0
û(k̂, r → ∞) ∼ cos(k̂r) + cot δ0 (k̂) sin(k̂r).

Let us also call v and v̂ two solutions at the same energies for the s-wave free
equation (that is, V = 0). We choose v in such a way that it is asymptotically
equal to u, that is

v(k, r) = cos(kr) + cot δ0 (k) sin(kr).

Notice that since V = 0, this corresponds to a irregular solution of the Schrödinger


equation. The same applies to v̂, which we denote as the (irregular) solution of
the free-particle equation that has the same large-r behaviour of û. In the limit
of k → 0 the behaviour of v̂ in the r → 0 limit is given by (1.48), which implies
r
v̂(r → 0) = v(k = 0, r → 0) = cos(kr) −
as
1.2. QUANTUM SCATTERING 25

Now let us consider the difference in the Wronskians for the two cases. In the
b → ∞ (upper) limit we have W (u, û) = W (v, v̂) because the two functions have
the same asymptotic condition. In the a → 0 (lower) limit we have W (u, û) = 0
because both u and û go to zero since they are regular solutions. In the same
limit, however, the v functions give
 
1
W (v, v̂) = 1 · k cot δ0 (k) − − .
as

Using the properties of v and v̂ discussed above, we have that in the limit a → 0
1
W (v, v̂) = −k cot δ0 (k) − ,
as

hence from W (u, û) − W (v, v̂) we obtain


Z ∞
1 2 1 reff
k cot δ0 (k) = − + k (vv̂ − uû) dr ∼ − + k 2 ,
as 0 as 2

where in the last equation we have expanded the right had side in powers of k and
we have defined the effective range as
Z ∞
reff = 2 (v̂ 2 − û2 ) dr.
0

Notice that we have chosen the same normalization throughout, that is by requir-
ing that the amplitude of the cosine quadrature be 1. ∗∗ In the effective range
limit, the differential cross section is still a constant (since we are considering
only s-wave scattering). It does, however, acquire a mild energy dependence. The
low-energy limit in this case can be thought of as kreff < 1.

Another important property of the phase shifts is that they are defined modulo π. This
ambiguity can be removed using the Levinson theorem, which relates the difference in the
phase shifts to the number of bound states. The theorem is easy proved by imagining of [Wei15,
enclosing our system in an impenetrable sphere with radius R  r0 , much larger than the §7.8]
range r0 of the potential. In this case the scattering solutions must go to zero at R, hence,
from Eq. (1.34) we must have that the wavevector k is quantized and satisfies

πl
kR − + δl (k) = nπ
2
so that the number of states with energy less than E (corresponding to a wavevector k
according to the usual E = ~k 2 /(2m)) is given by the integer part of
 
1 πl
N (E) = kR − + δl (k) .
π 2
∗∗
Ok, I guess the hard thing is to prove that the integral defining reff is indeed positive. I don’t know how
to do it except by hand-waving arguments like: û has to go to zero at the origin, while v̂ does not (it goes to
1, actually, so there is at least a region within the potential range where v̂ 2 > û2 . Since in the region where
the potential is negligible one has û = v̂, this shows that reff has at least the correct order of magnitude of the
size of the region where the effects of the potential are relevant.
26 CHAPTER 1. SCATTERING THEORY

The number of positive energy states in the presence of the potential is then

1
N+ (E) = N (E) − N (0) = (kR + δl (k) − δl (0)) .
π
One can use the same formula to calculate the number of energy states in the case of free
particles, where the phase shifts are zero, that is

kR
N+free (E) = ,
π
from which we can calculate the variation in the number positive energy states in the presence
of the potential as

1
∆N (∞) = N+ (E) − N+free (E) = [δl (∞) − δl (0)] .
π
So, where are the other states gone? They must have become bound (= negative-energy)
states once the potential has been turned on, hence the number of bound states nl is

1
nl = −∆N (∞) = [δl (0) − δl (∞)] . (1.55)
π
As mentioned, this result enables us to remove the ambiguity in the definition of the phase
shifts by requiring
δl (E → ∞) → 0,
hence the value of the phase shift in the limit of zero energy is equal to the number of bound
states (for that particular angular momentum).
It might be useful to check all these properties in an actual case. We show in Fig. 1.7 the
phase shifts of the first four angular momentum waves in the case of an attractive Gaussian
well, obtained from a numerical calculation. The behavior in the low-energy limits shows us
that there are two bound states with l = 0, one bound state for l = 1 and l = 2 and no bound
states from l = 2 onwards. At low energy the only contribution to the cross section comes
indeed from the s-wave: the calculated scattering length is as ∼ 1.72λ. In the region where
ka ∼ 2, we see that the l = 0 and l = 1 phase shifts are close to π. The contribution of these
waves is strongly suppressed, whereas the l = 2 shows a well pronounced resonant peak.

Exercise 4 Quantum cross section of the square well potential


Analyze the scattering of a quantum particle of mass m from a square-well potential, that
is a potential defined by 
−V0 r ≤ a
V (r) =
0 r>a
with V0 > 0.
Answer of exercise 4
The only parameter governing scattering in this system is

~2
ξ= ,
2ma2 V0
1.2. QUANTUM SCATTERING 27

Phase shifts /
2.00 l=0
1.75 l=1
l=2
1.50 l=3
1.25
1.00
0.75
0.50
0.25
0.00
0 2 4 6 8 10

Contribution to the cross section


3.0
2.5
2.0
1.5
1.0
0.5
0.0
0 2 4 6 8 10
ka

Figure 1.7: Phase shifts and contribution to the total cross section from the first four
partial waves in the case of an attractive Gaussian well V (r) = −V0 exp(−r2 /λ2 ), where
~2 /(2mλ2 V0 ) = 1/28.
28 CHAPTER 1. SCATTERING THEORY

after we use a as a unit of length, and V0 as the unit of energy. Inside the well one has a free
particle solution jl (k1 r) with wavevector
~2 k12 1
= E + V0 ⇒ (k1 a)2 = (ka)2 +
2m ξ
whereas we will denote by k the wavevector outside the well (that is, in the free-particle
region). Using continuity of the wavefunction and its derivative at r = a, and since for r > a
one has
ψ(r) = A (cos δl jl (kr) − sin δl nl (kr)) ,
we obtain
jl0 (ka) − Rjl (ka)
tan δl = (1.56)
n0l (ka) − Rnl (ka)
k1 jl0 (k1 a)
R ≡ , (1.57)
k jl (k1 a)
which is analytical, but not very much useful. Let us analyze the l = 0 case, which we know
should be dominant at small energies (that is, ka  1). In this case, it is more convenient to
consider the equation for u. In the case of s-wave scattering, the solution within the well is
sin(k1 x), whereas the one outside is A sin(kx + δ0 ). Since the function and its derivative have
to be continuous, one immediately obtains
 
k
δ0 = −ka + arctan tan(k1 a) .
k1
In the low-energy limit (ka  1) one can work out exactly the value of the phase shift, which
turns out to be h p  p i
tan δ0 = −ka 1 − ξ tan 1/ ξ
resulting in a negative scattering length, as expected, at least for large ξ. Notice that the
s-wave scattering length diverges when
1 π
√ = + nπ.
ξ 2
and the phase shift goes to π/2, which is the condition for the appearance of a resonance
(which we expect for ξ = 0.405, 0.045, . . .. To see what it means, let us recall the condition
for bound states in the square well. Here we must join the solutions A sin(k1 x) within the
well with Be−kx outside. The relation between k1 and k is
2ma2 V0
(k1 a)2 = − (ka)2
~2
which must of course be positive. Using the continuity condition gives the equation for the
energy levels
k
tan(k1 a) = −1
k1
that is, as a function of k
r 
ka 1 2
q tan − (ka) = −1.
1 2 ξ
ξ − (ka)
1.2. QUANTUM SCATTERING 29

A bound state is on threshold (that is, close to E ∼ 0) when ka → 0+ , which happens for
p  p 
ξ tan 1/ ξ → −∞

that is the same condition that results in s-wave resonance. In fact, the relation between
resonances at zero energy in s-wave scattering and the presence of quasi-bound states is a
general result known as Low theorem.

1.2.6 The Born approximation


[Wei15,
The calculation of the phase shifts δl (k) or the scattering amplitude f (k, θ) can be performed §7.4]
analytically only in very few cases. In general, one has either to resort to numerical calcula-
tions (easy, powerful, exact) or to approximations (analytical, you-know-what-you-are-doing,
rarely comparable to real-life situations). The most (mis)used approximation is named after
Born, and comes after a bit of contemplation of Eqs. (1.22) and (1.23): if the potential is
“small” (a precise meaning of which will be given below), then the scattered wave is only a
small perturbation on top of the incident one. If this is the case, then one can estimate f (k, θ)
0
by subsituting the incident wave eik·r in lieu of ψk (r0 ) in Eq. (1.23). This approach is known
as the Born approximation, and results in
m
Z
0
fB (k, θ) = − 2
V (r)ei(k−k )·r d3 r. (1.58)
2π~
Defining q = k − k0 and denoting by α the angle between q and r0 , the scattering amplitude
can be written as
m
Z
fB (k, θ) = − V (r)eiqr cos α 2πd cos α r2 dr
2π~2
m 2 sin(qr) 2
Z
= − 2 V (r) r dr
~ qr
2m
Z
= − 2 V (r) sin(qr) rdr
~ q
Z ∞
k
= − V (r)r sin(qr) dr. (1.59)
2E sin(θ/2) 0
where we have substituted q = 2k sin(θ/2).

Born approximation for the Yukawa and Coulomb potentials.. Eq. (1.59)
just begs to be calculated for a potential of the form
e−κr
V (r) = Ze2 ,
r
from which we get
k Ze2 q
f (k, θ) = − .
2E sin(θ/2) q 2 + κ2
This expression has a finite κ → 0 limit, which corresponds to the Coulomb
potential. The differential cross section turns out to be
Z 2 e4
σB = ,
16E 2 sin4 (θ/2)
30 CHAPTER 1. SCATTERING THEORY

which is exactly equal to the classical case (1.17). However, one must be aware
that the scattering theory that we are presenting is not expected to be valid in
the case of the Coulomb potential (why?). Luckily, the Coulomb potential is one
of the few fortunate cases that are analytically solvable in quantum mechanics:
the exact differential cross section turns out to be equal to the classical one (and
to the value obtained using the Born approximation).

It is not really straightforward to figure out the conditions of validity of the Born approx-
imation. Since it is derived by assuming that “the potential is a small perturbation”, one
expects it to be valid the the potential is “small”. There is only one other energy to compare
it with, which is the energy of the incoming beam, hence one expects the Born approximation
to become better as the energy of the incoming particles is increased. If the potential had a
typical energy scale V0 , the condition would then be E  V0 .
In general, this is exactly what is observed, but surprisingly the Born approximation turns
out to be valid in other circumstances, hence the “energy criterion” is not the most general. A
good indication of what this more general criterion could be comes from the optical theorem
(1.42). In the case of the Born approximation the scattering amplitude (1.59) is real, and
hence the optical theorem would imply a zero cross section, ΣB = 0, even though the cross
section calculated by integrating the square of (1.59) is certainly not zero. This apparent
contraddiction can be solved if we consider that the Born approximation is valid when it
produces a “small” cross section. Small, in this case, means smaller than the geometrical cross
section that can be estimated using a typical length scale of the interaction potential. If we call
this length scale a, then the condition of validity of the Born approximation becomes [Mes99,
XIX.§7, p. 812]
ΣB  4πa2 . (1.60)

In order to evaluate the total cross section in the Born approximation, let us rewrite (1.59)
using the variable q as

2m 1 ∞ 2m
Z
fB = − 2 V (r)r sin(qr) dr ≡ − 2 V(q),
~ q 0 ~

from which we get Z π


ΣB = |fB |2 2π sin θdθ.
0

If we now change variables from θ to q using q = 2k sin 2θ , one has

qdq = k 2 sin θdθ,

and hence
2k
8πm2
Z
ΣB = V 2 (q)qdq. (1.61)
~4 k 2 0

In the case of ka  1 the function V(q) can be evaluated to be of the order

V(q) ∼ V0 a3
1.2. QUANTUM SCATTERING 31

since sin(qr)/(qr) ∼ 1 in the region where V (r) 6= 0. In this case the Born cross section
becomes 2
m2 ma2

3 2 2
ΣB ∼ (V0 a ) k = V0 a2
ka1 ~4 k 2 ~2
which fulfills (1.60) when
~2
V0  when ka  1, (1.62)
ma2
that is, in the low-energy limit the Born approximation is valid when the potential energy is
much less than the kinetic energy of a particle confined in the region where the potential is
non zero. For an attractive potential this means that the potential should not have bound
states. Conversely, we can see what happens when ka  1, that is in the high-energy limit.
In this case, starting from the expression of V(q) does not really help because when k is large
q can still be small. One way is to start from expression (1.58) and work out the high-energy
limit; it can be done but it is mathematically quite challenging, so we refer the interested
reader to [Sch56]. Alternatively, one can go back to the Lippmann–Schwinger equation (1.20)
and consider the second term on the right-hand side (the scattered wavefunction) as small
compared to the first (the incoming wavefunction) when we subsitute the incoming wave to
ψk (r0 ). This derivation is in the spirit of “Born approximation works when the interaction
potential induces a small perturbation of the incoming wavefunction”. Taking into account
the fact that the scattered wavefunction has its maximum at r = 0, the condition that the
scattered wavefunction is just a perturbation of the incoming wave reads
Z
2m e ikr0
3 0 0
0 ik·r
d r V (r )e  1,
~2 4πr0

and the integration gives, assuming the V (r) has the shape of a potential well with magnitude
V0 and width a,
ikr eikr
Z Z 
2m 3 e ik·r
m 2 1  ikr −ikr
d r V (r)e , = r dr V (r) e −e
~2 4πr ~2 r ikr
Z a
m  
dr ei2kr − 1

= 2
|V0 |
~ k 0
m
i2ka

= |V | e − 1 − 2ika , (1.63)

0
2~2 k 2

for a potential of magnitude V0 that acts on a region of linear size a. Now we can take the
ka  1 limit, and the condition becomes, dropping all the numerical factors,

~2 k
V0  when ka  1. (1.64)
ma
Notice that in the low-energy limit, Eq. (1.63) reproduces the result of Eq. (1.64).

Born approximation on the phase shifts


As we have already discussed, at high angular momenta the centrifugal potential begins to
dominate, and hence the actual potential can be considered as a perturbation and the phase
shifts tend to zero. The actual values of δl for large l can then be effectively approximated
32 CHAPTER 1. SCATTERING THEORY

by Eq. (1.54) where we set V̂ = 0 and we consider ûl = uk = krjl (kr), obtaining the so called
Born approximation of the phase shifts
Z ∞
(B) 2m
δl = − 2 k V (r)jl2 (kr) r2 dr,
~ 0

which can be expected to be quite good for energies such that kr0 & l, where r0 is a typical
length of the potential.

1.2.7 The semiclassical approximation


[Wei15,
§5.7] Finding the phase shifts is easy numerically, but hard analytically hence one has to resort to
approximations. In this section we will derive some results using the (venerable) semiclassical
approximation of the Schrödinger equation. Semiclassical theory is explained very well in the
lecture notes by prof. C. Scrucca [Scr12]

The Wentzel–Kramers–Brillouin approximation


The Wentzel–Kramers–Brillouin (WKB) approximation starts from rewriting the time-independent
wavefunction as a function of the phase W (x), defined as
 
W (x)
ψ(x) = A exp i ,
~

which, after insertion in the Schrödinger equation, provides

1 i~ 2
|∇W |2 + V − E − ∇ W = 0, (1.65)
2m 2m
which just begs to look for an expansion of W (x) in powers of ~, that is

W (x) = W0 (x) + (−i~)W1 (x) + (−i~)2 W2 (x) + . . . (1.66)

from which one obtains, considering the terms up to O(~),

1
|∇W0 (x)|2 = E − V (x)
2m
1
∇W1 · ∇W0 = − ∇2 W0 .
2
Now,
p the first equation shows that we can identify ∇W0 with the classical momentum p(x) =
± 2m(E − V (x)), and hence write the solution
Z x
W0 (x) = ± p(x0 ) dx0 + const,

whereas if we stick to a 1D system, the second equation becomes


" !#
1 W 00 1 p 0 d 1
W10 = − 0
=− = ln p ,
2 W00 2p dx p(x)
1.2. QUANTUM SCATTERING 33

and hence we can write the semiclassical form of the solution of the Schrödinger equation as
i x
 
A
Z
0 0
ψsc (x) = p exp ± p(x )dx , (1.67)
p(x) ~

which makes a lot of physical sense. First, we see that the probability of finding the particle
in the region dx, |ψsc (x)|2 , is proportional to dx/p(x) ∝ dt, that is the time spent in that
region. Second, the behavior of the wavefunction is oscillatory in the classically allowed
region V (x) < E, and becomes exponentially (damped) in the classically forbidden regions
V (x) > E. Also, both signs in the exponential are allowed and due to the superposition
principle, the most general WKB solution of (1.65) is given by a linear combination of the +
and − sign in the phase factor of (1.67).
It is now a good time to look back at the derivation and trying to find out under which
conditions the expansion (1.66) is a good solution of the WKB equation (1.65). A first clue
is given by the form of the wavefunction in (1.67), where one can see that ψsc (x) → ∞ when
p(x) → 0, that is when one approaches the classical turning points defined by E = V (x). In
the other regions, one would expect the approximation to be valid as long as the O(~) term
in (1.65) is smaller than the square gradient term, that is when

~ ∇2 W (x)  |∇W (x)|2 ,


which means roughly


dp
~  p2 (x),
dx
or equivalently, a part from numerical factors O(1),

p2 (x)

~ dV
 , (1.68)
p(x) dx 2m
which means that the variation of the potential for one wavelength has to be smaller than
the kinetic energy. The physical meaning of this conditions can be grasped by assuming the
the potential V (x) has a typical magnitude V0 and a typical lengthscale a. In this case one
could estimate the “order of magnitude” of the force as
dV V0
∼ ,
dx a
and the order of magnitude of the kinetic energy as
p2 (x)
∼ V0 ,
2m
on account of the fact that the average kinetic energy must be of the same order of the
average potential energy (this is roughly the content of the virial theorem where, again, we
are neglecting O(1) factors). Introducing the de Broglie wavelength λ(x) = ~/p(x), the
condition of the validity of the WKB now becomes
λ(x)
 1,
a
which means that the wavefunction must have a large number of oscillations within the region
where the potential is different from zero. This is in accordance with the correspondence
34 CHAPTER 1. SCATTERING THEORY

principle which states that for large quantum numbers (that is, a large number of nodes in
the wavefunction) quantum mechanics becomes classical mechanics.
We have purposefully hidden the constants in Eq. (1.67), for their value depends on the
appropriate boundary conditions. Although we have seen that the solution (1.67) is not
expected to be valid near the classical turning points, there is a clever way to connect the
solutions on both sides of V (x) = E.†† In order to do so one has to analyze the solution of
the Schrödinger equation in a linear potential, because if V (x0 ) = E, then the potential can
be written as
V (x) = V (x0 ) + V 0 (x0 )(x − x0 ) = E − F (x − x0 ).

It turns out that the Schrödinger equation in that case can be solved analytically (invoking a
class of special functions named after Airy), enabling one to figure out the exact connection
between the wavefunctions on the opposite sides of a turning point. Let us assume that one
has a classically allowed region on the left of a turning point x0 , where the wavefunction is
written as

C0+ C0−
 Z x0   Z x0 
0 0 0 0
ψL (x) = p exp −i k(x )dx + p exp i k(x )dx ,
k(x) x k(x) x

where we have defined k(x) = p(x)/~, then the correct form of the wavefunction on the
right-hand side (the classically forbidden region) is
 Z x  Z x 
CA /2 0 0 CB 0 0
ψR (x) = p exp − β(x )dx + p exp β(x )dx ,
β(x) x0 β(x) x0

p
where we have defined β(x) = 2m(V (x) − E)/~ and the C coefficients are related by

C0+ = 1
eiπ/4 CA + e−iπ/4 CB 

2 (1.69)
C0− = 1
2 e−iπ/4 CA + eiπ/4 CB ,

or, equivalently,
CA = e−iπ/4 C0+ + eiπ/4 C0−
(1.70)
CB = eiπ/4 C0+ + e−iπ/4 C0−

Similarly, if the forbidden region is on the left-hand side of x0 , the two matching solutions
are
 Z x0  Z x0 
CA /2 0 0 CB 0 0
ψL = p exp − β(x )dx + p exp β(x )dx (1.71)
β(x) x β(x) x
C+ C−
 Z x   Z x 
ψR = p 0 exp −i k(x0 )dx0 + p 0 exp i k(x0 )dx0 , (1.72)
k(x) x0 k(x) x0

with the same relation between the C’s as above.


††
How far from the turning point the exact and approximate solution are similar enough is determined by
condition (1.68).
1.2. QUANTUM SCATTERING 35

Born–Sommerfeld quantization rule


Let us consider a potential well with classical turning points x0 < x1 . In zone I the wave-
function is  Z x0 
1 0 0
ψI (x) ∝ p exp − β(x )dx ,
β(x) x

so that we have CA = 1 and CB = 0. In zone II we have that the wavefunction is


  Z x1   Z x1 
1 iπ/4 −iJ 0 0 −iπ/4 iJ 0 0
ψII (x) ∝ p e e exp i k(x )dx + e e exp −i k(x )dx ,
k(x) x x

where we have denoted Z x1


J= k(x0 ) dx0
x0
so that we have, when considering the transformation to zone III,

C0+ = e−iπ/4 eiJ


C0− = eiπ/4 e−iJ .

Now, in zone III the wavefunction must be of the form


 Z x 
1 0 0
ψIII (x) ∝ p exp − β(x )dx
β(x) x1

so that we must have the condition CB = 0, that is


 
1
eiJ + e−iJ = 0 ⇒ J = n+ π
2

Semiclassical phase shifts


In the case of central potential, it turns out that the semiclassical approximation is more
accurate with the Langer form of the effective potential
2
~2 l + 12
Vcent (l, r) = ,
2mr2
which introduces a centrifugal barrier for all values of l, so that the boundary condition
u(r = 0) = 0 can be fulfilled.. Using the previous results for the WKB approximation, and
assuming an effective potential

Veff (l, r) = V (r) + Vcent (r)

that is monotonically decreasing and tends to zero at large separations we have just one
turning point r0 so that Veff (l, r0 ) = E and the left and right wavefunctions compatible with
the boundary conditions above are
 Z r0 
C 0 0
uL (r) = p exp − βeff (r )dr (1.73)
2 βeff (l, r) r
Z r 
C 0 0 π
uR (r) = p sin keff (r )dr + (1.74)
keff (r) r0 4
36 CHAPTER 1. SCATTERING THEORY

from which we read directly the expression of the semiclassical phase shift, that is the differ-
ence between the phase of the scattered wavefunction in the presence of the potential V and
the phase of the free wavefunction (which is obtained by letting V = 0). Hence

V r V
∞ Z ∞ 2
1 2m (l + 1/2)
Z
keff (r0 )dr0 =

δl = dr (E − V (r)) − , (1.75)

~ r0 ~2 r 2
r0 0

0

where the notation ()|V0 means that one has to take the difference of the integrals in (1.75)
with the interaction potential V (r) and in the free-particle case V (r) = 0. Notice that the
value of r0 depends whether the full potential or the free-particle case are considered.

Exercise 5 Semiclassical cross section of a hard sphere


What about the semiclassical approximation in the case of a hard sphere? How good do
you expect it to work?
Answer of exercise 5
Not very good. We know that at low energies the cross section is determined by the s-
wave scattering length. However, the semiclassical approximation, together with the Langer
prescription for the centrifugal potential, does not work very well for E → 0. In the case of
a hard sphere it is useful to use E0 = ~2 /(2ma2 ) as the unit of energy. With this definition
can see that if E = E/E0 < 14 one has r0 > a from which Eq. (1.75) would imply δl = 0. In
fact, one would get
r0 l+ 1
= √ 2,
a E
which can be equivalently written, noticing that E = ~k 2 /(2m) as

r0 l + 12
= .
a ka

Since the semiclassical phase shifts are zero unless r0 < a, we get that for a given value of ka,
there is a maximum angular momentum lmax = k − 12 . For these l ≤ lmax Eq. (1.75) becomes

 Z as
1 1 1
δl = l+ 2 − r 2 dr
2 r0 r0
  "s 2  2 #
1 a a
= l+ − 1 − arctan −1
2 r02 r02
  q  
1 2
E
= l+ E(l + 1/2) − 1 − arctan −1
2 (l + 1/2)2

the numerical solution of which shows that Σ(E  E0 ) → 2πa2 . This is the correct quantum
mechanical result. However, the fully quantum solution approaches it from above, whereas the
semiclassical one approaches it from below, as can be seen from the numerical result, plotted
in Fig. 1.8.
1.2. QUANTUM SCATTERING 37

Semiclassical cross section for a hard sphere


1.8

1.7

1.6
xs / πa2

1.5

1.4

1.3

1.2

0 25 50 75 100 125 150 175 200


E/E0

Figure 1.8: Cross section of the hard sphere potential in the semiclassical approximation.

1.2.8 Resonances
[Wei15,
§7.6] Phase shifts and resonance
As mentioned in the derivation of the relation between cross section and phase shifts, reso-
nances appear at energies where δl (E) = π/2. In this case we can expect that the l-th partial
wave gives the greater contribution to the amplitude (1.37) and, hence, the cross section.
From Eq. (1.37) we can then write

sin δl sin δl 1
fl (E) ∝ eiδl sin δl = −iδ
= = ,
e l cos δl − i sin δl cot δl − i

and then expand cot δl around δl (E0 ) = π/2,



d cot δl (E) 2
cot δl (E) = cot δl (E0 ) + (E − E0 ) ≡ (E − E0 ),
dE
E=E0 Γ

where the last line is the definition of Γ. We then get

1 Γ/2
fl ∝ = ,
(E − E0 )(2/Γ) − i (E − E0 ) − iΓ/2

whose imaginary part – proportional to sin2 δl , i.e. the contribution of the l-th wave to the
total cross section – is the venerable Breit–Wigner lineshape

Γ2 /4
sin2 δl (E) = . (1.76)
(E − E0 )2 + Γ2 /4

Semiclassical approach
The phase-shift approach to resonances is not particularly illuminating. One can derive the
from of the line, but it is not very clear what the meaning is. The semiclassical approach
38 CHAPTER 1. SCATTERING THEORY

Figure 1.9: Structure of the effective pair potential in the case of an attractive potential.
Taken from [Scr12].

enables us to understand the origin of resonances, linking the behaviour of the phase shift
and the lineshape to the appearance of quasi-bound states during the scattering process.
When the pair-potential is attractive, the effective potential may have for some specific
value of l the structure shown in figure 1.9. Using judiciously Eqs. (1.69) and (1.70) starting
from the “obvious” expression of the radial wavefunction in zone I, that is,
 Z r1 
C 0 0
uI (r) = p exp − β(r )dr ,
β(r) r

one can work out the expression of the wavefunction in the various zones. The solution in
zone II is
 Z x2   Z x2 
0 0 0 0
exp i k(x )dx exp −i k(x )dx
eiπ/4 −iJ x e−iπ/4 iJ
uII (r) = e p + e px ,
2 k(x) 2 k(x)

the one in zone III is


Z x3   Z x3 
0 0 0 0
exp β(x )dx exp − β(x )dx
sin J e−K x
uIII (r) = p + cos J eK px ,
2 β(x) β(x)

and one finally arrives at the following expression for the wavefunction in zone IV
    Z r 
1 −i π4 K i −K 0 0
uIV (r) ∝ p e cos Je + sin Je exp i k(r )dr +
k(r) 4 r3
   Z r 
i π4 K i −K 0 0
e cos Je − sin Je exp −i k(r )dr (1.77)
4 r3
Z r  
1 π i −2K
= p sin k(r) dr + + arg cot J + e , (1.78)
k(r) r3 4 4
1.2. QUANTUM SCATTERING 39

where one has


1 r2 p
Z
J(l, E) = 2m(E − Veff (l, r)) dr (1.79)
~ r1
1 r3 p
Z
K(l, E) = 2m(Veff (l, r) − E) dr, (1.80)
~ r2

and we have used the fact that cos(x − π/4) = sin(x + π/4). Comparison with Eq. (1.74)
shows that in this case the phase shift can be expressed as the sum

δl = δlnorm + δlres ,

where δlnorm has the form that it would have in the absence of a potential well (using the
outermost turning point r3 , cfr. Eq. (1.75)), whereas the resonant term
 
res i −2K
δl = arg cot J + e , (1.81)
4

is clearly related to the fact that an incoming particle can tunnel through the repulsive
centrifugal barrier between r2 and r3 and enters the potential well between r1 and r2 . For
some selected values of the energy E, the resonant contribution to the phase shift can become
the most important contribution to the cross section. For this to be the case one must have
δl ∼ π/2. Since the imaginary part of the argument of (1.81) is exponentially suppressed, in
order for the arg(·) function to be maximized one must have a very small real part. Hence
the energy and angular momentum must be such that cot J ∼ 0 which is equivalent to having
an energy E corresponding to that of a semiclassical quasi-bound state within the well, that
is  
1
J(l, E) = π n + . (1.82)
2
Given a solution E0 of the previous equation, assuming K(E) ∼ K(E0 ) and expanding
J(E) around this value,

d
cot J(E) (E − E0 ) = −J 0 (E0 )(E − E0 )

cot J(E) = cot J(E0 ) + (1.83)
dE E0

e−2K(E0 )
and defining Γ = − , one can write
2 J 0 (E0 )
 
res iΓ
δl = arg E − E0 + ,
2

which means
Γ/2
tan δlres = ,
E − E0
from which we see that the contribution to the total cross section around the energy E0 goes
like
Γ2 /4
sin2 δlres = , (1.84)
(E − E0 )2 + Γ2 /4
which is the venerable Breit–Wigner lineshape (1.76).
40 CHAPTER 1. SCATTERING THEORY

1.2.9 Coulomb scattering


[Wei15,
Coulomb potential is long-range. In the context of scattering, this means that it is eventually §7.9]
larger, in magnitude, than the centrifugal potential for any value of the angular momentum
l, because Coulomb is O(1/r) and the centrifugal potential is O(1/r2 ). Consequently, we
cannot assume that the scattering wavefunctions become eventually the superposition of free
spherical waves (that is, a superposition of jl (kr) and nl (kr)): the very concept of phase
shift looses its significance. Alternatively, the form (1.22) of the scattering wavefunction is
meaningless in the presence of the Coulomb potential.
Luckily, the Schrödinger equation in the presence of the Coulomb potentials is “exactly
solvable”, that is the solution is a particular confluent hypergeometric function (the so-called
Kummer function), which is not very informative. In the following, we will be content to just
state the solution, and point the interested reader to more appropriate references, such as
[Joa84, Chapter 6].
Before stating the main properties of this solution, let us try a heuristic approach. Since
for large r Coulomb potential is larger than any centrifugal potential, the large-r limit of the
Schrödinger equation for each angular momentum function is

~2 d2 ul Ze2
− + ul = Eul , (1.85)
2m dr2 r
which is conveniently written in reduced (atomic) units defining

~2
aB = (1.86)
mZe2
Z 2 e4 m
H0 = , (1.87)
~2
which are the (generalized) Bohr radius and Hartree energy, respectively. Introducing also
the reduced wavevector k̂ = kaB (with the usual relation between k and E, of course), the
equation becomes
d2 ul 2
− 2 + ul = k̂ 2 ul ,
dr r
which in the absence of Coulomb interaction would have the (outgoing) solution u(r) = eik̂r .
We can expect the presence of the long-range Coulomb potential to change the value of the
phase at large distances, that is we might expect a solution of the form
 
ul (r) = exp ik̂r + if (r)

with f (r → ∞) < k̂r because the Coulomb potential goes to zero, hence the gratest contribu-
tion to the phase must come from the free-particle equation. The equation for f (r) becomes
then
2
2k̂f 0 + f 02 − if 00 + = 0,
r
but given the properties of f (r) we might neglect for large distances the higher-order terms
(f 02 and f 00 ). We are then left with
df 1
=− ,
dr k̂r
1.2. QUANTUM SCATTERING 41

whose solution is
1
f (r) = − ln(2k̂r),

irrespectively of the value of the angular momentum.‡‡ This consideration immediately tells us
that the long-range nature of the Coulomb potential results in a modification of the stationary
scattering solution at large r, resulting in the appearence of an additional phase component
with logarithmic behavior.
When one uses the “analytic” solution, it is possible to write down the form of the scat-
tering wavefunction at large distances from the scattering center, in a manner analogous to
what has been derived in the case of the Lippmann–Schwinger equation (1.22) which enabled
us to define the scattering amplitude. The expression turns out to be

eikr−iξ ln(2kr)
ψ(r) = eik·r+iξ ln(kr−k·r) + fC (θ) , (1.88)
r
where we have defined
1
ξ =
kaB
Z ∞
Γ(z) = xz−1 e−x dx,
0

and the Coulomb scattering ampliude is written as


Γ(1 + iξ) −iξ log(sin2 (θ/2)) 2Ze2 m
fC (θ) = − e , (1.89)
Γ(1 − iξ) ~2 q 2

where, as usual q = 2k sin(θ/2). §§ The differential cross section is given again by the square
of the scattering amplitude (because the logarithmic part of the phase is negligible at long
distances) and we obtain
Z 2 e4
σC = |fC |2 = , (1.90)
16E 2 sin4 (θ/2)
that is the classical cross section, since |Γ(1 + iξ)| = |Γ(1 − iξ)|.

1.2.10 Scattering of indistinguishable particles


In am interacting two-particle system, one cannot really distinguish which particle is which
because of Heisenberg uncertainty principle. However, we use to describe this system with
a wavefunction Ψ(x, y) where we can in principle distinguish which particle is which by the
slot that has been assigned to its coordinate. That is, the value of Ψ(x, y) is the probability
amplitude of finding the first particle at x and the second one at point y. Since we cannot
really say which particle is where (we could in classical mechanics, because different particles
will have different and non-overlapping trajectories), we must have

Ψ(x, y) = eiϕ Ψ(y, x),


‡‡
The arbitrary integration constant has been chosen so that f (r) corresponds to the large-r limit of the
actual solution. I am not sure that there is a clever way to fix it.
§§
Notice that some times the logarithmic factor in Eq. (1.89) is not reported. Since it is a phase it disappear
when calculating the cross section. Notice, however, that in the case of indistinguishable particles (vide infra)
its presence does have a measurable influence.
42 CHAPTER 1. SCATTERING THEORY

that is, the probability of finding the first particle at x and the second at y must be the same
of finding the first at y and the second at y. Reiterating the indistinguishably condition, we
must have
Ψ(x, y) = ei2ϕ Ψ(x, y),
which has just to solutions ϕ = 0 (the wavefunction remains the same upon exchange of
arguments) or ϕ = π (the function changes sign upon exchange of arguments). In the first
case we deal with a symmetric wavefunction, in the second case we deal with a antisymmetric
one. In nature, the first case is observed with integer-spin system, whereas the latter is
observed in systems with half-integer spin. Luckily, there is also a theorem supporting this
observation (the famous spin-statistics theorem), but it can be proved in relativistic quantum
theory (a.k.a. Quantum Field Theory). We will of course not delve into that.
Nevertheless, we will see how this is reflected in our description of scattering. In our
approach so far, we have been using the center of mass R = (x + y)/2 and relative r = x − y
coordinate, where the wavefunction factorizes as
Ψ(x, y) = Φ(R)ψ(r).
Now, since the coordinate R is even upon particle exchange, the overall sign upon exchange
depends on the behavior of the function ψ(r) in relative coordinates: it is this part that can
be be even (ψ(r) = ψ(−r)) and odd (ψ(r) = −ψ(−r)). In order to reflect this in the case of
scattering process, we have to rewrite the Lippmann–Schwinger equation for the scattering
wavefunction (1.20) as the symmetrized form
  eikr
ψ(r) = eik·r ± e−ik·r + [f (k, θ) ± f (k, π − θ)] , (1.91)
r
from which is clear that the number detected particles in a solid angle dΩ is
~k
dn = |f (k, θ) ± f (k, π − θ)|2 dΩ,
m
but, what is the intensity that we need to calculate the differential cross section (1.1)? A
straightforward calculation of the relative current
~
jrel = Im [ψ ∗ ∇ψ] ,
m
associated to the source term ψ = eik·r ± e−ik·r would give zero, which is just as reason-


able because ψ represents a standing wave for the probability amplitude. Unfortunately, in
the case of indistinguishable particles, the analogy with scattering of classical waves is not
directly applicable. One should go back one step and think about the physical meaning of
the LS equation, which is in fact a stationary solution. The “stationary” source term is the
superposition of two waves with opposite wavevectors, which interfere in the scattering region
where the symmetry of the wavefunction has to be taken into account. Here we are dealing
with a limitation of time-independent scattering theory, which masks the actual boundary
condition which is of two (quasi) plane waves in the remote past impinging of one another.
The current we are interested in is therefore the relative current of the two components of the
stationary source term that we have to think as separated from one another (that is, when
the particles are so far away from each other that the symmetrization is immaterial), i.e.
   
~ −ik·r ik·r ~ ik·r −ik·r ~
I = Im e
∇e − Im e ∇e = 2k,
m m m
1.2. QUANTUM SCATTERING 43

which is twice the current of the “regular” LS equation for distinguishable particles (cfr.
Eq. (1.24)), and hence the differential cross section is

1
σ= |f (k, θ) ± f (k, π − θ)|2 . (1.92)
2
Notice the presence of the factor 1/2. In our case, this is consistent with the definition of
scattering cross section as proportional to the probability of a scattering event taking place.
On the other hand, if one thinks of a laboratory condition when a beam of particles is directed
onto a target made by the same particles and the scattered ones are observed, one has to keep
in mind that there will be two detections for each scattering event. Some authors prefer to
define the cross section as being proportional to the probability of observing a particle when
a beam of given intensity in shone on a target of like particles; in this case the cross section
would turn out to be twice as big as (1.92).
From a practical point of view, Eq. (1.92) has the consequence that for bosons only partial
waves with even angular momentum will contribute to the cross section, whereas for fermions
only odd angular momenta will.
Another important consequence of Eq. (1.92) is best seen when the observations are made
for θ = π/2, where the two amplitudes are the same. The differential cross section for bosons
in this case will be twice as the cross section that would be observed if the particles were
indistinguishable. In the case of fermions, the cross section would be zero, irrespectively of
the form of the potential.
The effects of bosonic or fermionic nature of scattering particles are particularly evident
in the case of Coulomb scattering, e.g. two electrons (fermion) or two α-particles (bosons).
Since the Coulomb amplitude is given by Eq. (1.89), the actual differential cross section for
two identical particles of charge Z will be given by

Z 4 e4 2 cos(ξ log cot2 (θ/2))


 
1 1
σCS (E, θ) = + ± , (1.93)
32E 2 sin4 (θ/2) cos4 (θ/2) sin2 (θ/2) cos2 (θ/2)

where the + sign is for bosons, and the − for fermions. This is what is indeed is observed, as
shown in Fig. 1.10. Equation (1.93) is known as the (nonrelativistic) Mott cross section.
Please bear in mind that the symmetry of the spatial part of the wavefunction is not
by itself related to the bosonic or fermionic nature of the particle. In fact, it is the full
wavefunction that changes or not sign upon exchange of particles. In general, when one
considers what happens when two particles are exchanged, there might be other quantum
numbers to consider apart from the position; spin being the typical example. In the case of
spin- 21 fermions, where the overall wavefunction must change sign upon exchange, this might
happen in two ways:

1. the spin part of the wavefunction is symmetric (triplet state) and the spatial part
antisymmetric

2. the spin part of the wavefunction is antisymmetric (singlet state) and the spatial part
is symmetric

so, if we have an unpolarized beam of electrons, the triplet state will happen with three-times
as much probability than the singlet, hence the total cross section will be a linear combination
of the even and odd cross section with weights 1/4 and 3/4, respectively.
44 CHAPTER 1. SCATTERING THEORY

Figure 1.10: Differential cross section for two α particles (charged bosons), from Ref. [HT56].
Notice that for θ = 90◦ the cross section is twice as much as the Rutherford one.
1.3. GENERAL SCATTERING THEORY 45

1.3 General scattering theory


In this section we will briefly present the general theory of scattering. This is a comprehensive
framework which is at the basis of the description of more complicated (that is, relativistic)
processes. The emphasis of this theory is on operators acting of free-particle states. It has,
however, an equivalent formulation in terms of the states themselves, from which we will
recover the results of the previous section. The description here follows the very good book
by Taylor [Tay06] to which the interested student is referenced for more details.
It is useful to recall some properties of scattering events in classical physics:
• Scattering trajectories have in and out asymptotes, which are free-particle solutions
(with positive energy)
• The in and out asymptotes characterize the trajectory, hence the scattering event
– however, for some potentials the particle can be trapped and fall into the center
(e.g., an attractive potential of the form −A/r2 )
• In general, there is a scattering trajectory for any asymptote (because we can evolve it
forwards and backwards in time)
• Not all trajectories evolve into asymptotes, because we might have bound states. But
any positive energy state at will evolve into an out asymptote and comes from an in
asymptote.

1.3.1 Møller operators and the S matrix


[Wei15,
Let us assume that we have a set of states that live in some Hilbert space H. The dynamics §8.1]
is H is described by an Hamiltonian H which can be written as

H = H0 + V,

where V denotes the interaction between the parts of H and H0 the free evolution. A scat-
tering process can be defined as a process in which an initial free state |ψin i evolves into a
state where the subsystems interact (|ψi, at time t = 0 say) and eventually evolves into a
final free-particle state |ψout i. The asymptotic “in” and “out” states should be eigenstates
of the free Hamiltonian H0 . In mathematical terms, if we denote by U (t) the time-evolution
operator associated to H and with U0 (t) the evolution operator associated to H0 , we should
have:

U (t)|ψi −→ U0 (t)|ψin i
t→−∞
U (t)|ψi −→ U0 (t)|ψout i,
t→+∞

which means that the time evolution at very early times (t → −∞) tends to the free evolu-
tion of some state |ψin i, the same happening at very late times where the evolution of the
interacting state is very similar to the evolution of some non-interacting state |ψout i.
We will assume that V is such that the asymptotic condition is verified, that is for every
|ψ0 i ∈ H there is a |ψi that has |ψ0 i as an ingoing asymptote, that is

U (t)|ψi −→ U0 (t)|ψ0 i,
t→−∞
46 CHAPTER 1. SCATTERING THEORY

Figure 1.11: Graphical representation of the Møller operators, taken from Ref. [Tay06].

the same happening for the outgoing asymptotes).¶¶ One can equivalently say that every
state in H is the asymptote of some trajectory. The converse is clearly not true in general:
if H has at least one bound state |φi then its evolution is U (t)|φi = eiϕ(t) |φi, and it cannot
represent the free evolution of anything for t → ±∞ (because the free evolution would spread
the wavefunction of any state). Under very general conditions one can show that the Hilbert
space of states can be decomposed as H = B ⊕ R, where B is the set of all bound states, and
R is the set of states which have both incoming and outgoing asymptotes. This property is
called asymptotic completeness: it is generally very hard to prove, and so we will not even
try to do that here.
However, this decomposition is full of interesting consequences. In particular for any
incoming state |ψin i we can write its t = 0 state as

|ψi = lim U † (t)U0 (t)|ψin i ≡ Ω+ |ψin i, (1.94)


t→−∞

the operator Ω− . Analogously, for every |ψout i ∈ H we can define the operator Ω+ as

|ψi = lim U † (t)U0 (t)|ψin i ≡ Ω− |ψout i. (1.95)


t→+∞

The operators Ω+ and Ω− , called Møller operators, connect all the asymptotic in and out
states, respectively, to the corresponding state at t = 0. We recall that these states are in
the subspace R which is orthogonal to the bound states of the Hamiltonian H in the Hilbert
space H. These properties of the Møller operators are summarized graphically in Fig. 1.11

Properties of the Møller operators. The mathematically inclined will have


noticed that the Møller operators, despite being the limit of unitary operators,
are not unitary in general. In fact, they connect the Hilbert space H with just a
subspace of it, that is the space orthogonal to the bound states, which we denoted
by R. They are unitary if and only if the potential does not have any bound
state.∗∗∗ Nevertheless, there are some properties that are still valid in general:
¶¶
For a spherically symmetric potential v(r) this is roughly equivalent to assume that v(r) falls off more
rapidly than r−3 at infinity, diverges less rapidly than r−2 in zero and has at most few discontinuities inbetween.
∗∗∗
Or if one works in a finite-dimensional space, which is definitely not our case.
1.3. GENERAL SCATTERING THEORY 47

• They preserve the norm, that is

k Ω± |ψi k=k |ψi k,

and for this they are called isometric.


• They are not invertible (unless they are unitary).
• Being isometric they have an adjoint that fulfills

Ω†± Ω± = 1,

but, in general, Ω± Ω†± 6= 1 unless they are unitary. In fact, one can see from
Fig. 1.11 that if one acts on a bound state with Ω† , then one does not come
back if Ω is applied to the result, because one would end up in one of the
scattering states.

One has that for every |ψi ∈ R multiplying Eqs. (1.94) and (1.95) by Ω†−

|ψout i = Ω†− Ω+ |ψin i ≡ S|ψin i, (1.96)

which defines the operator S (usually called, S matrix). Notice that S connects the whole H
to itself, and is norm conserving. This is enough to conclude that S is unitary. It takes an
asymptotic “incoming” state and produces the corresponding asymptotic “outgoing” state.
A very important property of the S matrix is the so-called intertwining relation, which states

HΩ± = Ω± H0 . (1.97)

In order to prove it, one can write

eiHτ Ω± = eiHτ lim eiHt e−iH0 t


t→∓∞

= lim eiH(t+τ ) e−iH0 (t+τ ) eiH0 τ


t→∓∞
= Ω± eiH0 τ ,

and differentiate with respect to τ setting then τ = 0. One important consequence of the
intertwining relation (and its hermitean conjugate) is

SH0 = Ω†− Ω+ H0 = Ω†− HΩ+ = H0 Ω†− Ω+ = H0 S,

that is
[S, H0 ] = 0, (1.98)
which means that the in and out states must have the same energy.

1.3.2 The S matrix: cross sections, transition probabilities and all that
[Wei15,
Let us for the time being consider the case of two-body elastic scattering. As we now, the in §8.2]
and out states are eigenstates of momentum, hence the matrix elements of S are hp0 |S|pi. If
there is no scattering S = 1 (identity), so S = 1 + R. In general,

hp0 |S|pi = δ3 (p0 − p) − 2πi δ(Ep − Ep0 )t(p0 , p), (1.99)


48 CHAPTER 1. SCATTERING THEORY

where t is the on-shell T -matrix, because it is defined for |p0 |2 = |p|2 . The factor −2πi is
introduced for later convenience. Also one can define the scattering amplitude as

f (p0 , p) = −~(2π)2 m t(p0 , p), (1.100)

or
i
hp0 |S|pi = δ3 (p0 − p) + δ(Ep − Ep0 )f (p0 , p). (1.101)
2πm~
Notice that according to this definition f (p0 , p) has the physical dimensions of a length.

S matrix and cross section

Let us now derive the relation between the S matrix and the differential cross section. We will
start again from the general relation (1.1). If we consider sharp momentum eigenstates, the
corresponding intensity I is given by the (uniform) current associated to the sharp momentum
state
1
ψ(x) = hx|pi = eip·x/~ , (1.102)
(2π~)3/2

that is
~ 1 p
I= Im [ψ ∗ (x)∇ψ(x)] = 3
. (1.103)
m (2π~) m

Now we have to evaluate the number of particles per unit time scattered in a given solid angle.
The off-diagonal part of |S|2 gives the transition probability (proportional to time, due to the
presence of the square of a δ function as discussed in appendix 1.B) to a specific direction
determined by the vector p0 . To calculate the number of particles in a solid angle, we have
to multiply it by the volume element d3 p0 and perform the integration over the modulus
(remember, we still have a δ function of the energy to use). Hence

|hp0 |S|pi|2 3 0
dn = d p
T
1 T 2
= 4π 2 δ(Ep − Ep0 )|t(p0 , p)|2 p0 dp0 dΩ
T 2π~
2π 2
= δ(Ep − Ep0 )p0 dp0 |t(p0 , p)|2 dΩ
~

= mp |t(p0 , p)|2 dΩ, (1.104)
~

so that from Eqs. (1.103) and (1.104) we have

1 dn
σ =
I dΩ
= (2π)4 ~2 m2 |t(p0 , p)|2
= |f (p0 , p)|2 ,

where we have used the definition (1.100).


1.3. GENERAL SCATTERING THEORY 49

S matrix for general processes


In the previous sections we have using as examples for the S matrix calculations the non-
relativistic elastic scattering process discussed above. However, the formalism is not limited to
this simple case, since the Hilbert spaces for the “in” and “out” states can really be anything,
usually multi-particle states. Since we are not dealing with relativistic processes involving
the creation and destruction of particles the “in” and “out” spaces must be the same space.
This restriction can however be lifted (keeping mostly the same formalism!) by considering
H as a Fock space in Quantum Field Theory.
Nevertheless, let us briefly discuss the form of the S matrix in the case of multiparticle
scattering. Let us denote as α and β all the quantum numbers describing states in H. In
this case we have both the conservation of energy and total momentum, so that the matrix
elements of S can be written as

hβ|S|αi = Sβα = δ(β − α) + δ(Eβ − Eα )δ (3) (Pβ − Pα )Mβα , (1.105)

where we expect M to be a continuous function of its parameters. Taking the square of Sβα
(let us consider for the moment that β 6= α, which is the typical case for scattering) we have
to deal again with the squares of the δ functions. In the case of energy, we can again use
Eq. (1.116), whereas in the case of momentum, analogous considerations enables us to write
h i2 V
δ (3) (Pβ − Pα ) = δ (3) (Pβ − Pα ),
(2π~)3
where V denotes the volume of the box where the process described by S takes place. It
is understood the one has to take the V → ∞ limit at the end of the calculation. Since
we are working in a finite-volume box we should, for consistency, also use a finite-volume
normalization of the free-particle states, that is
eip·x/~
ψp (x) = hx|pi = √ ,
V
so that one can write for the transition probability per unit time between α and β
N +N −1
1 (2π~)3 α β

(box) 2
Γ(α → β) = |Sαβ | = δ (3) (Pβ − Pα )δ(Eβ − Eα )|Mβα |2 (1.106)
2π~ V
where the matrix elements of M are evaluated with the infinite-volume normalized states
(1.102), which are much more easy to work with.
We can now generalize the definition (1.1) of the differential cross section to more general
processes. In all cases, one measures the number of particles per unit time scattered in
some configuration, hence we need to sum the transition probability over the set of state
corresponding to the measured configuration. Q Since the ending configuration is determined
3
by the momenta of the final particles dβ = d pβ and the density of states in momentum
space is V /(2π~)3 , the sum over the final states over some range produces the differential
transition rate
 Nβ Y
V
dn = Γ(α → β) d3 pβ (1.107)
(2π~)3
N −1
1 (2π~)3 α

= |Mβα |2 δ (3) (Pβ − Pα )δ(Eβ − Eα )dβ. (1.108)
2π~ V
50 CHAPTER 1. SCATTERING THEORY

Integration of the transition rate over all possibile outcomes produces the total rate,
N −1 Z
1 (2π~)3 α

Γα = |Mβα |2 dΞ, (1.109)
2π~ V

where we have defined the final state space

dΞ = δ (3) (Pβ − Pα )δ(Eβ − Eα )dβ, (1.110)

which depends both on the energy-momentum relation (classical or relativistic) and the num-
ber of outgoing particles. For Pα = 0 and Eα = E (which is ok for both the decay of a
single particle or the scattering of two particles in the center-of-mass frame), and assuming
non-relativistic particles one has

p21 p2
Eβ = + 2 ,
2m1 2m2
and
p21 p2
  
(3)
dΞ = δ (p1 + p2 )δ E − + 2 d3 p1 d3 p2
2m1 2m2
p21
 
= δ E− p21 dp1 dΩ
2m

= m 2mE dΩ = mp1 dΩ,

where we have introduced the reduced mass m = m1 m2 /(m1 + m2 ). Recalling the general
definition (1.1) with
p1 1
I= ,
mV
for the finite-box momentum eigenstates, we recover the expression for the cross section in
the center-of-mass frame for the α → β process
Z
σβα = (2πm~)2 |Mβα |2 dΩ. (1.111)
1.3. GENERAL SCATTERING THEORY 51

Appendices of Chapter 1
1.A The Wronskian theorem
This is a very useful theorem, connecting integral properties of the solution certain types of
linear differential equations (which include Schrödinger equation) in one dimension. Here I
follow the presentation of Messiah [Mes99, Sec. III.§8]. Let us begin by defining the Wronskian
of two functions y1 (x) and y2 (x) as

W (y1 , y2 ) = y1 y20 − y10 y2 . (1.112)

The Wronskian theorem asserts that, given two solutions y1 and y2 of ordinary differential
equations of the form

y100 + F1 (x)y1 = 0 (1.113)


y200 + F2 (x)y2 = 0 (1.114)

with a “well behaved” pair of functions Fi (x) (= piecewise continuous), the overall variation
of the Wronskian in given by
Z b
W (y1 , y2 )|ba = [F1 (x) − F2 (x)] y1 (x)y2 (x) dx. (1.115)
a

The proof is straightforward. Let us multiply the first equation by y2 , the second by y1
and subtract. We get a term of the form

d
y2 y100 − y1 y200 = − W (y1 , y2 ),
dx
which is just the opposite of the derivative of W (y1 , y2 ). Integration in the interval [a, b] then
proves the theorem.

1.B Squares of δ functions


The presence of δ functions in the expression for S (1.99) might at first sight appear troubling
because we recall that S enables the calculation of probability amplitudes. The actual prob-
abilities are squares of matrix elements of S, such as |hβ|S|αi|2 . These will involve squares
of δ functions, which might seem not so well defined. To see the physical meaning of these
infinite terms, let us recall the definition of the delta function:
T /2
1
Z
δ(x − y) = lim ei(x−y)t dt,
2π T →∞ −T /2

which is of course a limit. Using this expression we might write


T /2
1 T
Z
2
[δ(x − y)] = lim ei(x−y)t dtδ(x − y) ∼ δ(x − y), (1.116)
2π T →∞ −T /2 2π

with the understanding that we must take the T → ∞ limit at the end of the calculation.
Don’t be fooled by the names of the variables so far. However, let us consider the cases of
52 CHAPTER 1. SCATTERING THEORY

energy and momentum, for they are clearly appearing in (1.99). For the energy we might
write Z T /2
1 T
δ(Eβ − Eα ) = ei(Eβ −Eα )t/~ dt ∼ , (1.117)
2π~ −T /2 2π~
when the last equality is meant to be valid the the square of a δ function involving energy
is required. In the previous equation, T has indeed the dimension of time. Hence transi-
tion probabilities proportional to the square of a δ function are proportional to transition
probabilities per unit time.
Bibliography

[BDT+ 76] D. Bassi, M. G. Dondi, F. Tommasini, F. Torello, and U. Valbusa. H-Ar potential
from high-resolution differential cross-section measurements at thermal energy.
Phys. Rev. A, 13:584–594, 1976.

[Chi96] M. S. Child. Molecular Collision Theory. Dover, unabridged corrected edition,


1996.

[HT56] N. P. Heydenburg and G. M. Temmer. Alpha-alpha scattering at low energies.


Phys. Rev., 104:123–134, Oct 1956.

[Joa84] C. J. Joachain. Quantum Collision Theory. Elsevier, 1984.

[Mes99] A. Messiah. Quantum Mechanics. Dover, 1999. Two volumes bound as one.

[Sch56] L. I. Schiff. Approximation method for high-energy potential scattering. Phys.


Rev., 103:443–453, Jul 1956.

[Scr12] C. Scrucca. Quantum Physics III. Institute for Theoretical Physics, Lausanne
(Switzerland), 2012.
http://documents.epfl.ch/users/s/sc/scrucca/www/notes/qpIII.pdf.

[Tay06] J. R. Taylor. Scattering Theory. The Quantum Theory of Nonrelativistic Colli-


sions. Dover, 2006.

[TWW79] J.P. Toennies, W. Welz, and G. Wolf. Molecular beam scattering studies of orbiting
resonances and the determination of van der Waals potentials for H–Ne, Ar, Kr,
and Xe and for H2 –Ar, Kr, and Xe. J. Chem. Phys., 71(2):614–642, 1979.

[Wei15] S. Weinberg. Lectures on Quantum Mechanics. Cambridge University Press, sec-


ond edition, 2015.

53

Вам также может понравиться